基于国家急诊CT影像数据库的多病种精准快速联合筛查的数学方法与系统

快速、准确地筛查急诊多器官的多重损伤，对于把握创伤救治的黄金时间、降低患者致残率及致死率具有重要意义。本项目将构建高效急诊CT筛查辅助诊断系统，用以解决急诊影像数据量庞大、影像解读复杂和损伤机制认知不完善的临床困境。项目组建立全国多中心健康人群CT数据库，基于前期深度学习分割模型构建万例健康人群多器官标签数据库和急诊创伤CT多器官标签数据库；在异源多中心CT数据的图像规范化重建基础上，基于图像特征先验以及多器官位置关联，构建数学模型驱动的多器官分割策略；将三维融合信息引入泛化连接函数的多分类广义线性模型中，构建多器官空间位置约束下的多疾病分类模型；利用贝叶斯网络构建多器官与多疾病的损伤机制，从而实现后验概率的快速计算，以辅助急诊影像的精准筛查；集成部署基于先验知识和空间约束的多器官分割和多疾病分类模型，搭建全流程“一查多病”急诊智能诊疗平台,并开展多中心应用验证，普惠我国民众的医疗健康。

It is significant to screen multi-organ injuries in the emergency department, which can decrease the time to diagnosis and expedite treatment, then decrease the mortality and disability rate consequent to trauma. This project will establish an efficient diagnosis system for emergency CT screening to solve the clinical dilemma of massive emergency image data, complex interpretation, and incomplete cognition of multi-organ injuries mechanism. Our group will build a national multicenter CT database of healthy people, then construct annotated CT database containing multi-organ labels of 10,000 healthy people and emergency trauma patients based on the previous deep learning segmentation model. The standard reconstruction of heterogeneous multicenter CT image data will be set up, and a multi-organ segmentation strategy driven by mathematical models will be developed based on image feature priors and multi-organ location association. The 3D fused information will be introduced into the multi-classification generalized linear model with a generalized connection function, and the multi-disease classification model will be constructed with multi-organ spatial location constraints. The Bayesian network is used to demonstrate the multi-organ and multi-disease injury mechanism to realize the rapid calculation of posterior probability and to assist in accurately screening emergency images. The multi-organ segmentation and multi-disease classification models will be integrated and deployed based on prior knowledge and spatial constraints, and a whole-process "one examination of multiple diseases" emergency intelligent diagnosis and treatment platform will be built. Then multicenter application verification will be carried out to benefit the medical health of Chinese people.